Abstract:OBJECTIVE To establish a model for predicting smoking rates for secondary students in Anyang according the individual factors and health behavior. METHODS To investigate 2160students in 8secondary school in Anyang from Jan. 2013 to Dec. 2013 and establish a model for predicting smoking rates by getting basic information, health related behavior and other factors according to the questionnaire of Chinese adolescent health risk related behavior issued by the Chinese Center for Disease Control and Prevention for the different types of schools. RESULTS The multivariate analysis screened out 19 factors that were significant for the model. The information applies to the establishment of Logit model ( =18.383, P=0. 343). It had significant difference for 22 factors in the Logit model affected the response variables in null hypothesis (Wald = 162.37). It is showed that the model can preferably fit smoking rates(GFI=293.28,dispersion degree=0.27, P=1.000)and had no excessive dispersion (dispersion degree＜1). The results showed that it can be applied to sub model prediction (R=0.896，R2=0.803) and the identified models are reliable (F=11.239, P＜0.05) . The results of the adjusted R2 for 2142 data of the cumulative model was less than the original adjusted R2 showed that fitted Logit cumulative model has some practicability. The predicted value and the true value had more consistency for different students. CONCLUSION Smoking of the secondary students can be predicted by establishing a credible model and it can be reduced smoking rate and intervening smoking behavior among adolescent by the risk factors of smoking of the secondary students.